How Ethereum devs will use AI to boost blockchain’s development

GM, Tim here.

One of the most influential voices in the Ethereum development community wants the $238 billion blockchain to lean hard into artificial intelligence.

Tomasz Stańczak, co-director of the Ethereum Foundation, argues that the large language models that power AI can deliver big efficiency gains by writing, reviewing and editing proposals, moderating meetings, writing code, and even accepting or rejecting network upgrades.

“It is important for Ethereum to be the first chain to be LLM-driven,” Stańczak said on Sunday. “It is an advantage akin to being the first PoW chain.”

PoW, or Proof-of-Work, refers to a consensus mechanism blockchains like Bitcoin use to validate the network.

Ethereum has a big head start, Stańczak said. Much of the blockchain’s development, in the form of written proposals, developer calls, and other technical documentation, is published online. An AI could easily be trained on this trove of information, giving it deep insight into the entire history of Ethereum’s development.

It’s a bold idea, and likely one of Stańczak’s last initiatives before he leaves the Ethereum Foundation at the end of the month.

Software developers outside of the crypto industry increasingly rely on AI.

Last week, Spotify co-CEO Gustav Söderström said in his firm’s fourth-quarter earnings call that the firm’s top developers “have not written a single line of code” in 2026, instead using AI exclusively to design software.

Stańczak’s ideas go beyond using AI to just generate code.

He wants to bring in AI agents, autonomous software programmes designed to achieve specific goals without constant human oversight.

The agents could participate in the submission and editing of Ethereum Improvement Proposals, moderate developer calls in real-time and provide suggestions to participants, and eventually distill and broadcast the entire Ethereum governance process.

Outside of crypto, businesses are pouring resources into developing AI agents, with tech giants Google and Microsoft leading the charge.

The market for AI agents is estimated to surpass $50 billion within the next five years, according to Boston Consulting Group.

It all sounds very futuristic, akin to video game AI companions like Cortana in the Halo franchise.

But the buzzy technology is not without drawbacks.

AI models are prone to hallucinating — a term used for when AI produces false or made up information that it tries to pass off as real.

In April, research conducted by OpenAI found that its two latest and most powerful reasoning models hallucinated 33% and 48% of the time.

These hallucinations can be disastrous for AI agents. Those who have tried getting them to engage in dynamic, fast-paced activities like crypto trading have found they often make simple mistakes.

Ethereum watchers may need to wait a while before they start seeing AI make its way into the blockchain’s development.

Stańczak said he expects the shift to take at least two years.

“We should have the relevant tooling and integrations finished by Q3 (the earlier the better),” he said.

Top DeFi stories of the week

This week in DeFi governance

VOTE: ZKsync Token Assembly to fund bug bounty programme on Immunefi

VOTE: Compound DAO to approve new Ecosystem Protection and Continuity Fund

VOTE: Fluid to use treasury funds to cover user losses caused by ETH borrow rate spikes

Post of the week

Sam Bankman-Fried is trying to convince Crypto X that FTX was never insolvent, despite serving a 25-year prison sentence for defrauding his customers.

It’s not going well.

Tim Craig is DL News’ Edinburgh-based DeFi Correspondent. Reach out with tips at [email protected].